Power Flow Control for Power and Voltage Management in
Future Smart Energy Communities
A Fazeli†, M Sumner*, E Christopher*, M Johnson*
* University of Nottingham, Nottingham, NG7 2RD, UK, mark.sumner@nottingham.ac.uk
† Alstom Grid, St Leonard’s Avenue, Stafford, ST17 4LX, UK
Keywords: DSM, Smart Grid, Voltage Regulation.
Abstract
The Community Power Flow Control (CPFC) Algorithm has
been proposed as a technique for managing electrical power
and energy within small communities. The CPFC manages
the resources in the community (DSM, energy storage, RES)
in order to control the community’s instantaneous power flow
according to a target set by a higher-level management
system. This paper investigates the capability of the CPFC to
manage local distribution voltage levels. The power flow data
for a community has been used together with transformer and
cable impedance models to demonstrate the variation of the
distribution voltage in a community with significant
penetration of PV and EV. The CPFC is shown to manage the
voltage levels along a feeder within the statutory limits, even
when there is significant export.
1 Introduction
The electrification of heat and transport in addition to the
uptake of Distributed Generation (DG) in the residential
sector are expected to affect the operation of the distribution
networks, and would not be possible unless the required
degree of active management and control is incorporated
within this system. In particular the integration of these
emerging resources is expected to cause voltage instability for
the radial low voltage distribution network.
Voltage stability can be defined as the ability of a power
system to maintain steady and acceptable voltage at all the
buses in the system at normal operating conditions, after
being subjected to a disturbance [1]. Voltage instability can
occur when a disturbance causes a sudden and uncontrollable
decline in voltage. There are various factors which could
cause voltage instability including:
• Contingencies (line or generator outage due to faults)
• Load disturbances
• Improper operation of voltage control devices
• System faults
In the absence of voltage compensation devices, the reduction
of voltage as a result of load increase would eventually reach
a critical point, beyond which any further load increase would
lead to a complete voltage collapse and a blackout [2]. As
1
continuous and uninterrupted delivery of power to consumers
is the vital functionality of electric power systems, different
regulatory limits for variation of voltage and frequency have
been put in place. In particular for the UK low voltage
network the voltage statutory limits are +10% and -6% of the
230 Volts nominal voltage [3].
2. The Community Power Flow Controller
The Community Power Flow Control Algorithm is part of a
hierarchical approach to electrical power and energy
management proposed in [4,5] and illustrated in Fig.1.a. The
CPFC algorithm manages the instantaneous power drawn
from the distribution network by the energy community
according to a Community Power Flow Target (CPFT) passed
to it from a higher level of the hierarchical smart grid control
framework. The CPFC manages all of the energy resources
within the community including embedded storage (fridges,
heat pumps and related thermal stores such Direct Heat Water
Cylinders (DHWC)), time-displaceable loads (washing
machines, dryers, dishwashers, electric vehicles (assuming
NO vehicle to grid (V2G))) and explicit electrical energy
storage such as the battery, V2G etc. The CPFC optimises the
use of these resources on a second by second basis according
to the decision making flowchart of Fig. 1.b [5] to ensure that
the community follows as closely as possible the CPFT. It can
thus serve as the basis for a Virtual Power Plant. The CPFT
itself is generated by the higher level energy management
algorithm - the Community Energy Profiler (CEP) - and its
profile is continually changed according to an optimisation
function, which at present aims to minimise the community's
energy cost, self-consumption, peak current drawn from the
network or a weighted combination of all three. The CEP can
use information from the network operators (for example
Time of Use tariffs), and from neighbouring communities to
determine the CPFT, together with any prediction of the
available local resources (PV, wind, microCHP) and
community loading for a particular time period
(day/week/peak period).
The hierarchical control structure for optimal dispatch of
Distributed Energy Resources (DERs) within the distribution
network appears as a realistic method of incorporating active
resources within the distribution network. The proposed
deterministic optimization and coordination algorithm has
been shown to be a suitable method for real time dispatch of
resources at the community cell level of the hierarchy [5].
The cooperative optimal dispatch of cold loads and
community energy storage unit has shown the effectiveness of
cooperative resource dispatch, towards meeting the common
objective function of maintaining community's power flow
below a pre-specified threshold.
Fig.1a Hierarchical Control Framework
Fig.1b. Hierarchical Control Framework
In order to investigate the impact of different penetration
levels of EV and DG on voltage stability within the low
voltage distribution network, and also examine the
effectiveness of the CPFC algorithm at indirectly regulating
2
the voltage at different nodes, a model of the LV network
supplying a community of 100 dwellings has been created.
For this work the CPFT is set at a constant level each day
which is the average net consumption predicted for the
community based on average consumption and average
renewable energy source (RES) production each day. For this
work, a year’s work of community loading has been extracted
from the electricity demand model at [6], and PV data is
available from local measurements with 5 min resolution,
providing community equipment usage and power flow data
which can be optimised by the CPFT. The structured
management of the resources is able to control the CPF close
to its CPFT, and in particular significantly reduces the
evening peaks [5].
3. Modelling a Small Distribution System
Load flow analysis is the process of calculating the steady
state voltage magnitude and phase angle at every bus and the
power flow in each branch of the distribution network given
the status of generators and loads. The voltage is determined
by solving a set of nonlinear equations representing the net
active and reactive powers flowing out of each bus. Because
these equations exhibit nonlinear characteristics, they must be
solved iteratively [7].
It is worth mentioning, that having studied different load flow
analysis techniques including Newton Raphson, Gauss-Seidel
and Kersting’s Ladder Iterative [8], it became apparent that
such methods have different characteristics in terms of
applicability to different network topologies and simulation
speed, and hence are suitable for different purposes. For
example an ant colony optimisation algorithm has been
developed in [9] and used to perform power flow analysis for
a transmission network with FACTs devices. It reported that
since operation of FACT devices is introduced into the
problem the resultant power balance equations are much more
non-linear. However the developed power flow technique
presented in [9] is capable of solving the power balance
equations for that system. A modified version of the forward
propagation method has been presented in [10]. [10] applies
this method to a radial distribution network with distributed
generation, and has demonstrated the accuracy and
convergence speed of the proposed method.
Before explaining the adopted modelling approach in this
project, it is necessary to demonstrate the selected network
topology and the methodology used for calculating the
impedances of this network’s cables.
3.1 The network topology and parameters
Fig. 2 shows the schematic of this candidate network,
connecting a secondary transformer to 100 dwellings. The
total feeder impedance is effectively the summation of the
impedances connecting different buses, which are assumed to
be of equal length. The resistor represents the total active
power in each dwelling. As the distribution of power over the
low voltage network occurs over shorter distances in
comparison with the transmission network, the low voltage
distribution networks are mostly resistive [11], [12].
Fig. 2 Schematic of the candidate network, connecting a secondary
transformer to 100 dwellings
Therefore the feeder inductance is assumed to be zero. The
impedance of the LV feeder depends on its cable
characteristics (cross section, material), in addition to its
overall length. Since the feeder length would determine its
overall resistance and the resultant voltage drop across the
feeder, it is important to determine this parameter with respect
to the maximum tolerated voltage drop across the feeder
length. Therefore the maximum feeder length is calculated so
that the voltage drop at the feeder end node does not fall
below the lower statutory limit of 216.2Vrms. This
calculation includes the After Diversity Maximum Demand
(ADMD) which represents the average power per dwelling,
and the resistivity of the feeder conductor. A three phase
185mm2, mains LV cable with an aluminium core has been
selected from [13]. The value of ADMD obtained from [14] is
2kW and the maximum feeder length is calculated as 3.07km.
This feeder length has a total feeder resistance of 0.47Ω.
Assuming that the dwellings are equally distanced from one
another, every feeder section between different dwellings has
been assumed to have a resistance of 0.0047Ω. Having
calculated the total feeder length and impedance for the
network under investigation, Kersting’s ladder iterative load
flow analysis technique has been used to determine the
voltage variation at every bus across the feeder with a time
resolution of 1 second.
The ladder iterative technique is an algorithm developed
specifically to compute power flow for radial networks. This
method relies heavily on application of Kirchhoff’s Current
Law (KCL) and Kirchhoff’s Voltage Law (KVL) at all
connection buses across the network, in a series of forward
and backward sweeps in order to converge to a solution [15].
This method is essentially comprised of two iterative
calculation stages termed the forward and the backward
sweeps in which the line currents and bus voltages are
respectively determined. For the network topology under
consideration shown in Fig.2, the forward sweeps starts from
the feeder end point (i.e. bus 100). In order to start the first
iteration the voltage is assumed to be equal to the reference
voltage of 1p.u. The value of the load power is also known at
every bus. It is then possible to calculate the line current in
each bus.
Having calculated the bus currents and assuming a value of
1p.u. for the voltage at bus 100, the voltage at the adjacent
bus can be calculated by applying KVL. These steps are
consecutively applied between adjacent nodes in order to
3
calculate the line currents, until the second bus is reached. At
this point the last calculated line current is compared with its
previous value. If the difference between the two line current
values is less than the convergence tolerance (10-6) the loop
terminates, otherwise the backward sweep is executed. The
backward sweep starts at bus 2, where the value of the line
current between bus 2 and bus 3 calculated in the forward
sweep is used to determine a new value for voltage at bus 3.
Then the load current is simply calculated by applying Ohm’s
law and the application of KCL at bus 3 gives the line current
between bus 3 and bus 4. The backward sweep continues until
the last bus voltage (i.e. V100) is calculated. Upon termination
of the backward sweep, the forward sweep is initiated using
the most recently calculated value of V100 as an input. The
execution of both the forward and the backward sweeps
continues until the voltage variation at all the buses converges
to a value less than the specified tolerance.
After performing the Ladder Iterative load flow analysis
technique on the network topology shown in Fig. 2, the
voltage magnitude at every bus across the feeder is calculated.
Fig. 3 shows the voltage variation obtained at the100 different
buses, with an approximate drop of 3% at bus 100.
Fig. 3. Voltage variation at different buses across the feeder with 1kW
load in every dwelling
An entire day’s worth of simulation using the Ladder Iterative
technique takes 18 minutes 43 seconds which is an acceptable
simulation speed for this work.
4 Simulation Scenarios
As highlighted before, the electrification of heat and transport
in addition to the integration of distributed generation, are
envisaged to occur in the near future. However the integration
of these emerging load and generation units are expected to
affect the operation of the distribution network, and would not
be possible unless the required degree of active management
and control is incorporated at different levels of the system. In
particular significant voltage variation beyond the statutory
limits is expected for the radial low voltage distribution
networks which could consequently lead to voltage
instability.
4.1 The Load Model
In order to investigate the impact of different penetration
levels of EV and DG on the low voltage distribution network,
and also examine the effectiveness of the CPFC algorithm at node with 30% EV, 50% GSHP and PV penetration levels. It
indirectly regulating the voltage at different nodes, a model of is worth noting that in order to investigate the worst case
the LV network supplying a community of 100 dwellings has scenario, the allocation of these resources starts from the
been created as described in [5]. The model runs a real time dwelling at the feeder end node.
simulation of 100 different dwellings. The electricity demand
model presented in [6] is used to quantify the variation of
electrical appliance load throughout the day. Bespoke time
domain models for each of these pieces of equipment are used
to obtain a year’s worth of loading data for each simulated
house, sampled at 1 minute intervals. The simulation also
includes LI-Ion battery, Photovoltaic Panels, Ground Source
Heat Pumps and Electric Vehicle models. The CPFC
algorithm can interact with equipment to shed or enforce
connect suitable devices based on the requirement to follow
the community power flow target. The community power
flow is quantified by adding the power from different DERs
(i.e. battery, PV, GSHP and EV) and the appliance load from Fig. 4. Voltage variation at three nodes across the feeder for a
community of 100 dwellings with no EV or DG
all households.
According to Fig. 5. when there is no control of active power
The electrical loads and DERs from each of the 100 house- (red trace), the voltage magnitude falls below the lower
models is connected to one of the nodes of the distribution statutory limit repeatedly (shown by the green dotted line of
system shown in Fig 2 to demonstrate the variation of voltage Fig 5.). This is caused as a result of EV charging during the
along the radial feeder with time during the day. evening hours. The voltage can reach a value as low as 200V,
if there is no control of active power.
4.2 Voltage Variation with No EV or PV
The application of the CPFC algorithm significantly reduces
The voltage drop across the LV feeder is directly proportional the occurrence of voltage excursions below the lower
to the feeder length and the distance from the secondary statutory limit (blue trace). However minor voltage
transformer. Therefore while the voltage at the connection excursions still occur during the interval 23:00 - 04:00 hours
node between the secondary transformer and the LV feeder of the second day. This is primarily due to simultaneous
could be effectively regulated close to the nominal value, the reconnection of electric vehicles for charging – all are
voltage at the feeder end node would suffer the most configured for overnight charging.
significant variation. This effect will be more profound when
there is either an increase in load or distributed generation, as
both cases entail an increase in feeder current and
consequently a significant voltage drop across the feeder.

This point is illustrated in Fig. 4 which shows the voltage

variation at three different equally distanced nodes across the
feeder. The voltage variation at the different nodes are
obtained from the LV network model (with no electric
vehicles or embedded PV generation), in which the Ladder
Iterative power flow analysis technique is performed at every
second during a period of 1 day and 9 hours. Node one in Fig.
4. represents the feeder connection point to the secondary
transformer. Node two is the feeder middle point and node n
is the feeder end point.
According to Fig. 4. the voltage at the end node suffers the
most significant reduction, as a result of significant voltage
drop across the feeder length.
4.3 Load Growth with and without CPFC
In order to study the effect of load growth on voltage
variation at the LV level, three different penetrations of
electric vehicles within a community of 100 dwellings are
modelled and the simulation results are presented in this
section. Fig.5 shows the variation of voltage at the feeder end
Fig. 5. Variation of voltage at the end of feeder with 30% EV
. Fig.6 and Fig.7 show the variation of voltage at the feeder
end node for the same community this time with 60% and
90% EV penetration levels respectively. The magnitude of
voltage excursion in the absence of the CPF control increases
to 192V (60% EV penetration) and 197V (90% EV
penetration) as shown in Fig.6. and Fig.7. This could lead to
instability, and consequently voltage collapse. On the other
hand the application of the CPFC algorithm effectively
regulates the voltage during the first day for both the 60% and
the 90% EV penetration levels. However as shown on Fig.6.
at 60% EV penetration there are still significant voltage
excursions during the interval 02:30 and 05:15 hours of the
second day. These are due to the combination of over
shedding the GSHP and EV loads, which is also alleviated

4
when the EV penetration is increased to 90% as shown on
Fig.7. According to Fig.7 the regulated voltage with the
CPFC and 90% EV penetration drops as low as 201 V, which
might also cause instability.
Fig.6. Variation of voltage at the end of feeder with 60% EV
Fig.7. Variation of voltage at the end of feeder with 90% EV
It is worth noting that at 90% EV penetration during the
interval 00:00 – 07:00 hours of the first day, the GSHP are
enforce connected to increase the CPF to a value very close to
its target. However since voltage is not used in the CPFC’s
decision making this enforce connection operation leads to
minor voltage excursions below the lower statutory limit.
Since the magnitude of these excursions is relatively low,
they can be ignored.
In order to demonstrate that the incorporation of sufficient
energy storage could improve the CPFC’s effectiveness at
indirectly regulating the voltage, the same community is
allocated with a 150kWh battery and the EVs are assigned
high shedding flexibility. Fig.8. shows the voltage variation at
the feeder end node for this community. According to Fig.8 it
is evident that as the voltage approaches the lower statutory
limit during the late evening hours of day one and the
morning hours of day two, only minor voltage excursions
occur, and effective voltage regulation is achieved indirectly
by implementing the CPFC algorithm.
It is worth noting that the location of the connection node of
the battery to the main feeder in addition to the battery’s
power rating would significantly influence voltage regulation
and these parameters should be taken into account at the
design stage.
5
Fig 8.Variation of voltage at the end of feeder with 90% EV including
community battery
4.4 DG growth with and without CPFC
Increasing the penetration level of distributed generation at
the low voltage level could result in reverse power flows,
during periods when the load on the network is low and there
is peak generation. In that case the feeder current flows from
the dwellings towards the secondary transformer, and voltage
increases across the feeder length as we move away from the
secondary transformer towards the feeder end node.
Consequently the occurrence of voltage excursions beyond
the upper statutory limit (10% nominal voltage) at different
nodes across the feeder is expected. The feeder end node
would be experiencing the most severe over-voltages. In
order to study this effect, three different penetrations of PV
panels (1.68kWp per house) within a community of 100
dwellings are modelled and the simulation results are
presented in this section. The penetration of other DERs is
kept at a constant level in order to merely capture the
influence of distributed generation on the voltage variation.
Fig.6.38 shows the voltage variation at the feeder end node
for three different PV penetration levels. It is evident that
during the period 11:00 – 17:00 hours, when irradiation and
consequently PV generation is at its highest level, voltage at
the feeder end node rises above the 230V nominal value.
However the voltage rise at 30% and 60% PV penetration are
still below the upper statutory limit during the peak
generation period. Despite that at 90% PV penetration the
voltage reaches the upper statutory limit and during the
interval 11:00 – 13:00 hours it occasionally crosses the upper
limit by up to 2V.
In this study it was assumed that the voltage at the secondary
transformer (node one) is constant at 230 volts. In addition to
that the selected PV units had relatively modest power
ratings. If however the voltage at the transformer was at a
higher level, or if the rated power of the PV units is greater
than 1.68 kW then more severe voltage excursions beyond the
upper limit are expected even at 30% and 60% PV
penetration. Fig.9 shows the voltage variation at the feeder
end node for the same community at 90% PV penetration,
both with and without CPF control.

Variation of voltage at the end of feeder at different PV penetration
levels
Fig 9. Voltage variation at the feeder end node at 90% PV penetration
It is evident that application of the CPFC algorithm has
indirectly resulted in effective voltage regulation very close to
the nominal value of 230 volts throughout the day. The CPFC
algorithm has effectively curtailed the correct magnitude of
PV generation, in order to control the CPF at its target value.
This real time control of active power has indirectly regulated
the voltage at different nodes across the feeder.
It is however important to note that the DG curtailment
considered in this study would not be beneficial for the
consumer. On the other hand if a bulk battery with sufficient
capacity were available within this community, the locally
generated energy would be stored in the battery and not
curtailed. In that case both the consumer and the DNO would
benefit from the operation of the CPFC algorithm and the
battery. That justifies the need for energy storage at
community level.
5 Conclusion
The impact of load and DG growth on voltage variation
within a candidate section of the LV network supplying a
community of 100 dwellings has been investigated. The
simulation results show that, when there is no control of
active power, even at 30% EV penetration the voltage
magnitude falls below the lower statutory limit repeatedly and
can reach a value as low as 200V. On the other hand the
application of the CPFC algorithm significantly reduces the
occurrence of voltage excursions below the lower statutory
limit. However since the CPFC algorithm does not use
voltage as a control parameter it fails to regulate the voltage
at 60% and 90% EV penetration levels, as more significant
voltage excursions occur. This could lead to instability, and
consequently voltage collapse. The impact of power
generation from PV units is not as severe and only minor
voltage excursion above the upper voltage statutory limits are
observed even at 90% PV penetration. However it is
important to note that the impact of DG on voltage might be
more significant if DG units with higher rated power were
considered. In conclusion, the inclusion of voltage as a
secondary control parameter in the CPFC’s decision making
could ensure effective control of power and voltage
throughout the day. This could also avoid unnecessary DG
curtailment and has been suggested as future work from this
project.
Acknowledgements
This work was supported by European Regional Development
Fund (ERDF), within the Accelerating Low Carbon Economy
(ALCE) project.
References
[1] P. Kundur, Power System Stability and Control, 1994.
[2] F. Castellanos and A. F. Mohammed, "Voltage stability and reactive
power compensation on the T&TEC system," in SoutheastCon, 2003.
Proceedings. IEEE, 2003, pp. 134-139.
[3] I. Grau, L. M. Cipcigan, N. Jenkins, and P. Papadopoulos, "Microgrid
intentional islanding for network emergencies," in Universities Power
Engineering Conference (UPEC), 2009 Proceedings of the 44th
International, 2009, pp. 1-5.
[4] Fazeli, A.; Christopher, E.; Johnson, C. M.; Gillott, M.; Sumner, M.,
"Investigating the effects of dynamic demand side management within
intelligent Smart Energy communities of future decentralized power
system," Innovative Smart Grid Technologies (ISGT Europe), 2011
2nd IEEE PES International Conference and Exhibition on , vol., no.,
pp.1,8, 5-7 Dec. 2011
[5] A. Fazeli , Sumner, M, Johnson, C M, Christopher, E,, "Coordinated
Optimal Dispatch of Distributed Energy Resources within a Smart
Energy Com-munity Cell," presented at the 2012 3rd IEEE PES
Innovative Smart Grid Technologies Europe Berlin, 2012.
[6] I. Richardson, M. Thomson, D. Infield, and C. Clifford, "Domestic
electricity use: A high-resolution energy demand model," Energy and
Buildings, vol. 42, pp. 1878-1887, 2010.
[7] P. J. Lagace, M. H. Vuong, and I. Kamwa, "Improving power flow
convergence by Newton Raphson with a Levenberg-Marquardt
method," in Power and Energy Society General Meeting - Conversion
and Delivery of Electrical Energy in the 21st Century, 2008 IEEE,
2008, pp. 1-6.
[8] I. Richardson, "Integrated high-resolution modelling of domestic
electricity demand and low voltage electricity distribution networks,"
PhD, Electrical and electronic engineering, Loughborough, 2011.
[9] Y. Kumar, K. Dwivedi, and G. Agnihotri, "Development of Ant
Algorithm for load flow analysis," in Power Systems Conference and
Exposition, 2009. PSCE '09. IEEE/PES, 2009, pp. 1-5.
[10] D. Bhujel, B. Adhikary, and A. K. Mishra, "A Load Flow Algorithm
for Radial Distribution System with Distributed Generation," in
Sustainable Energy Technologies (ICSET), 2012 IEEE Third
International Conference on, 2012, pp. 375-380.
[11] T. L. Vandoorn, J. D. M. De Kooning, B. Meersman, and L.
Vandevelde, "Soft curtailment for voltage limiting in low-voltage
networks through reactive or active power droops," in Energy
Conference and Exhibition (ENERGYCON), 2012 IEEE International,
2012, pp. 545-550.
[12] Y. Vencken, W. Labeeuw, and G. Deconinck, "Dealing with an
overdose of photovoltaics at distribution level," in PowerTech
(POWERTECH), 2013 IEEE Grenoble, 2013, pp. 1-5.
[13] eon, "Cables, Cable Laying & Accessories Manual " 2006.
[14] e. o. c. networks, "Network Design Manual."
[15] G. T. H. Alejandro D. Domínguez-García, Siddharth Suryanarayanan,
"Implications of the Smart Grid Initiative on Distribution
Engineering," University of Illinois at Urbana/Champaign2011.